Methodology

Stumpage’s public price history comes from MI DNR stumpage price reports. Searchable report history begins in October 1994; coverage varies by reporting period, species and geography. Quarterly reports on file begin in October 2006; the oldest history uses twelve-month reporting windows. Source files are fetched directly from the DNR’s public server and stored with three things attached — the address it came from, the date we fetched it, and a SHA-256 fingerprint of the exact file. That is the receipt. If the DNR’s copy and ours ever diverge, the fingerprint says so.

These are sales of public timber — standing timber the DNR sells from Michigan’s state forest land. On private land, stumpage prices can run higher or lower, and private-land sales are not in this record; that gap is exactly why the contributed private record sits alongside the public one, in its own visual language, never blended into it.

Beside the price reports sits a second public record: the DNR’s bid-opening results since fiscal 2004 — recorded sales, bidders and per-species offers. The two do not always agree, and the difference is the DNR’s, not ours. Since 2017 the price report’s per-species sold price is a contract price: when a winning bid is skewed — one line bid far above its appraisal — the DNR cuts that line and spreads the difference across the sale’s other species, leaving the sale total unchanged. The bid record keeps what was actually bid. Stumpage shows each figure under the source that produced it and never blends the two. Where a sale’s two figures disagree, the receipt behind each one names the document it came from, so the difference is always traceable to its source rather than to us.

We read price lines from the DNR’s available spreadsheets and CSV files. PDF copies are retained as source documents but are not read into the searchable price data. A price line summarizes sales for a species, product, unit, geography and reporting period; it is not an individual transaction price. At last count the record holds 665 source documents and 164,568 price lines, across 306 reporting periods, covering Oct 1, 1994 through Jun 30, 2026.

Map boundaries are the DNR’s volume determination units as published; the display copy is simplified to about 5 m, and each sale’s own sketch on Openings uses the verbatim geometry.

Once per quarterly edition Stumpage issues a market letter — the edition in brief, built from this same record, rendered to a PDF, fingerprinted and kept with its edition in the Record. Subscribers receive it by email.

When Stumpage says red oak sawtimber averaged $187 per MBF (illustrative) in a period, that is a volume-weighted average, not a simple one. A sale of 40,000 board feet counts for forty times a sale of 1,000 — big sales pull the average, exactly as they do in the real market. A plain average would let one tiny outlier distort the picture; a volume-weighted average reflects where the wood actually traded.

The range beside every average is just as literal. The minimum and maximum are the lowest and highest prices in real recorded sales for that species, product, and period — observed, never modeled, never a confidence interval. And a price is never converted from one unit to another. An MBF (a thousand board feet) and a cord are different physical things: sawtimber sells by the MBF, pulpwood by the cord. A price always carries the unit it sold in, and we never restate it in another.

Prices now ranks the edition’s statewide cells — one per species, product and unit — by their volume-weighted average sold price, highest first. A ranking never crosses a unit: pulpwood sold by the cord and by the ton make two lists. What moved compares each cell with the same period a year earlier (four quarterly reports back) and shows the percent change in that average price, or in the volume sold in the unit of sale. Only cells with at least 10 reported sales in both periods qualify, and Prices now states that floor beside each list. A cell absent from either period is a gap, not a zero.

The Michigan Timber Index summarizes changes in reported Michigan DNR stumpage prices. Separate indexes track sawtimber and pulpwood.

Each period we use species weights fixed from the configured base window, separately for each product and native unit. A species missing from that period is omitted from both the current-cost sum and the base-cost sum; its price is never filled in. We compare the cost of the available members with their average cost across the base window. That baseline is set to 100. An index of 112 means the basket costs 12% more than in that stretch; an index of 94 means 6% less. This Laspeyres calculation holds base weights fixed, so changes in current sold volumes do not reweight it. Available basket members can change between periods, however; missing members can affect comparisons.

In one line: the latest reported cost of the available basket members, divided by those members’ base-window cost, times 100.

The configured base is January 2021 through December 2025. It is fixed and reassessed deliberately, not rolled forward automatically each year. Base weights use eligible observations in that window, including overlapping quarterly, six-, nine- and twelve-month reports. Those observations are report lines, not independent transactions. Every index figure and receipt names its stored base window. Today that window is Jan 1, 2021 to Dec 31, 2025.

The DNR changed its sawtimber product classification in 2017. The modern sawtimber index therefore does not combine observations from before and after that change. Each of its windows begins where the record does: mid-2016 for the twelve-month window, and 2017 Q2 for the quarterly one. Pulpwood carries no such break, and its index is published as far back as the DNR reported the window — to 1994 for the twelve-month window, and to 2006 Q4 for the quarterly one, which is the first quarterly stumpage price report on file. So a pre-2016 index value exists for pulpwood and never for sawtimber, and every series names the earliest window it holds rather than asking a reader to remember which.

The index is a ratio, not a price. It is published by quarter and, since 1995 for pulpwood and 2017 for sawtimber, by calendar year.

The Prices page compares annual changes in the Michigan Timber Index with changes in Midwest consumer prices. Both lines are set to 100 in the first year of the window; the published index value, with its own base window named beside it, travels with every point so the figure still traces. Consumer prices are the BLS Consumer Price Index for All Urban Consumers, All items, Midwest region, not seasonally adjusted (series CUUR0200SA0, 1982–84 = 100) — Michigan sits inside that region, and a not-seasonally-adjusted CPI is final when issued. A year’s value is the mean of its published months; at least 11 of 12 must exist (BLS never published October 2025) and the count is stated, and fewer is a gap, never a fill.

real_t = nominal_t ÷ (CPI_t ÷ CPI_reference), the optional inflation-adjusted line, where CPI_reference is the mean of the published monthly values across the index’s own base window. The annual sawtimber index begins with 2017 because the DNR renamed its product code that year; pulpwood runs from 1995.

Optional context lines (housing starts, lumber and pulp price indexes, diesel, the DNR’s own auction measures) are averaged by calendar year and set to 100 in the same first year, share the axis for comparison only, and keep their raw values and units in the tooltip and table. They describe conditions; co-movement does not prove causation, and the events on the timeline provide context without attributing any index move. Every term used here is defined in the data dictionary.

One product at a time — sawtimber and pulpwood are never blended or converted. The event notes on the chart mark documented changes in the record or the market (a 2017 report renames sawlogs to sawtimber and changes its averaging) and are never offered as causes.

When no usable price is available for a species and period, Stumpage shows a gap — a break in the line, labeled “no data reported.” A gap may reflect no reported sales, missing source coverage or withheld observations; it does not by itself establish that no timber sold. We never interpolate across it or guess what the missing number might have been.

The DNR’s published reports occasionally contain data-entry errors — lines no Michigan timber sale could have produced. We checked the worst of them against the original files: the impossible numbers sit verbatim in the DNR’s own documents (a $5,000-per-cord pulpwood line; a line whose own average falls outside its own min–max range). These errors appear in the original source files. Affected rows are excluded from published calculations and retained in the archived data with an explanation.

So there is one documented rule. A reported line is withheld from every served price, range band, distribution, and the index when it: (1) exceeds a per-product ceiling far beyond any credible sale — pulpwood above $500 a cord, sawtimber above $8,000 per MBF, or an average above $200 / $3,000 respectively; (2) contradicts itself — a minimum above its maximum, or an average outside its own min–max range beyond rounding; or (3) prices a product in a unit that product does not trade in — pulpwood in board feet, sawtimber in cords.

We never alter a number. Exclusion is all-or-nothing per line, the excluded lines stay in the raw record with the rule that caught them visible, and any chart missing withheld lines says so in plain text. As of Aug 2026 this withholds 1,072 of 164,568 reported lines — about 0.7%. Unusual-but-possible prices are not excluded: a startling number that could be real stays in the record.

Charts cap their price axis near the top of the average line so the average stays readable when a few period ranges peak far higher; the tooltip and the table under each chart carry every full range. A chart that omits withheld lines says so beneath it and links here.

One wrinkle in the source data: the DNR reports the same sales over overlapping windows — 3-, 6-, 9-, and 12-month spans that share underlying sales. Stumpage treats the quarterly (3-month) window as the canonical read, so each sale is counted once and periods line up cleanly.

The longer windows are shown in the Record as the DNR published them. After a bid opening the DNR may restate a quarter’s lines in a later report; the Record keeps every edition, and a receipt names the edition its figure came from.

The geography levels overlap. The DNR reports a sale at every geography it belongs to: a sale in the Gwinn Management Unit is also a sale in the Escanaba State Forest, in the Upper Peninsula and in the statewide record. Stumpage aggregates each level from the lines the DNR reported for that level, so a management unit’s price lines are counted again in its state forest, its DNR region and the statewide row. The four levels are four readings of one record, not parts of a whole, and their counts do not sum: seventeen management units hold 88,328 price lines between them where the statewide row holds 12,696. Comparing a unit with the state is sound; adding them is not.

Comparisons align one species, product and native unit across two to four reporting regions on identical calendar quarters or calendar years. Annual comparisons use January–December windows and exclude overlapping rolling reports. History spans end with the latest calendar period on record; five years starts January 1 of the fifth year, including a partial final year. Year-over-year change compares the same calendar period a year earlier. Missing prices break the line.

Reported sale counts are source aggregates, not unique sales. A report line is a species/product/unit/geography observation in a report, not a transaction. Evidence counts usable lines and distinct source documents within the selected reporting dates. Coverage is the union of reported months, so overlapping windows do not inflate it. The active view identifies missing periods and excluded observations; its source details retain precise period receipts, document retrieval dates and the latest usable observation.

Distribution percentiles describe report-line average prices, with each line one observation; they are not percentiles of individual sale prices. The median divides those observations in half, the middle half runs from the 25th to 75th percentile, and the outer marks show the 10th and 90th percentiles. The volume-weighted average is a separate reference.

Annual volume uses the DNR’s January–December reports, never the sum of overlapping reporting windows. Values retain their native units and separate product panels. Appraised volume sold is the timber quantity in the sale record; it is not harvested volume or immediately available supply. Missing years stay empty, and an incomplete or stopped record is labeled beside the result.

Reported sale-count sums add the report’s species lines; a sale containing several species can contribute to several lines. They are not unique sales. The DNR’s 2017 sawlogs-to-sawtimber reporting change remains marked in the history rather than presented as a continuous classification.

Reporting geographies overlap: a management unit, its state forest, its DNR region and the statewide row describe different levels of the same record. Compare them without adding their totals. The current map excludes retired management-unit boundaries; their observations remain accessible in the archive and regional histories. Map boundaries do not turn counties, offices or buying territories into price-report regions.

Sale history describes each recorded bid offering. A winning total is the sum of line bids times the DNR’s estimated native quantities; premium over the advertised minimum is that winning total divided by the advertised total, minus one. An explicit no-bid result differs from an opening whose outcome is not established. Offered counts include those states; sold figures use the recorded sold results.

Bidder win rate is offerings won divided by offerings bid on; winning value sums offers on won offerings. Normalized company spellings are grouped with their original DNR spellings available for inspection. Market-history filters that do not apply to a chart remain named beside it. County acreage and preparation costs have separate counting rules below.

Today provides quick views; the pipeline joins prospectus offerings, bid results, revision notices and the dated unsold-list snapshot for investigation. An event is keyed by the exact sale number and opening date. An undated listing stays undated. Event and listing counts are not unique sales, harvested timber or available inventory. Reoffer means an earlier opening date exists for that exact sale number and can overlap other statuses.

A prospectus opening on or after the view’s as-of date with no overriding outcome is scheduled. A past date without a joined result is “result not in record”; it does not establish that the opening occurred or remained unsold. Sold, no bids and withdrawn labels require the corresponding recorded evidence. A dated unsold snapshot may disagree with later outcomes; both remain visible. Its report date and source retrieval date are separate, and neither establishes current availability.

Pipeline dates, geography and search apply before the activity counts; the activity filter then selects the rows. Filters intersect, while month and status groups cover only the current page. A recent unsold snapshot can include an older or undated listing in a recent window. The unsold sheet’s field office may differ from the office that held the bid opening, so each is named separately where recorded.

Today’s quarter view keeps the current incomplete quarter separate from complete comparison quarters. Its volumes are offered/sold estimates in the unit of sale. On Openings, recent winning prices are volume-weighted for the same species, product and native unit; the statewide and office lookback windows are stated in the active view.

The reader retains native line quantities and advertised unit prices. The DNR’s whole-sale statistical cords estimate is identified separately; Stumpage does not convert or sum across units. Sale conditions, operational requirements and shared bid instructions remain substantive document content.

Citations use physical, one-based PDF pages. A shared table page is cited once, while Sources & notes keeps field-level mappings. A document link without a page means a precise location was not established. “Not extracted” does not mean the source omitted the field. Failed extraction, unavailable archived files, unlocated sections and pages without usable text are distinct states. No map interpretation or optical character recognition is inferred. Load archived PDF switches page links to the authorized local copy; Open prospectus retains the original source.

A territory is a saved union of office, management-unit and county selections. Overlapping selections do not duplicate events; additional pipeline filters intersect with that union. Offices are matched by their recorded labels, not assumed county boundaries. A county is not a price-report region; same-name regional links offer context, not territory totals.

County or unit geography from a related offering’s prospectus is approximate and excluded by default. Opting in keeps “Approximate” beside the matching row. Suffix-based associations use a unique earlier-or-same-date prospectus and never merge offering histories or prove a reoffer. Future prospectuses cannot supply geography to earlier events. Saved territory links use the latest saved selections; workspace links reopen the current record rather than a frozen snapshot.

Buyer Analysis describes public DNR bid-opening evidence, separately from the longer aggregated price-report archive. One buyer at one opening is one participation, even when the document repeats species lines or raw name spellings. Buyers found only in a highest-bidder header are retained with that evidence type. Stable name-group IDs survive source-row reloads. Punctuation, case, accents and legal suffixes follow the existing deterministic normalization; a normalized name is not proof of a verified legal entity.

The default “Recently observed: last 24 months” window ends at the latest usable completed opening in the loaded record. Twelve-month, five-fiscal-year and all-history windows are also available. Five fiscal years include the fiscal year containing the cutoff date, with each year running October 1 through September 30. An explicit historical as-of excludes later opening events. It selects event dates in today’s archive, not an archived version of what Stumpage knew then. Source retrieval, latest observed opening and analysis dates are separate.

Global activity is derived from participation, not wins, and stays independent of profile geography. “No recent DNR bid observed” says nothing about whether a business is operating. Operating, closed, acquired or unknown status needs its own dated business evidence. Saved relative windows follow the current record; an explicit as-of remains fixed. Neither freezes the underlying sources.

Known-outcome win rate is recorded wins divided by known outcomes. Unresolved results, conflicting winners and ties are shown separately; a missing result is never a loss. Reoffers remain separate opening events in one sale history. A later offering after a win does not establish default, contract nonexecution or completed harvest.

Species and product choices come from the observed offering catalog for the selected period and geography. Choosing a species narrows the product and volume-unit choices. Species composition uses the full offered basket, keeping each recorded species, product, volume unit and bid classification together. Participation share divides matching buyer participations by all of that buyer’s participations. Opportunity share divides those matching participations by matching observed offerings in the same period and geography. Multi-species shares can sum above 100%. Won volume is recorded estimated offered volume on won openings, not actual removals or deliveries; different volume units never convert or combine.

Office shares use nonoverlapping recorded offices. Own geographic mix divides office participations by the buyer’s selected participations; share of known sold openings divides buyer wins by all known sold openings in that office and period. Volume shares use separate species/product/native-unit denominators. These are shares of the observed DNR record, not all timber purchasing. County filters default to exact prospectus matches and report missing coverage. An explicit earlier-offering geography option includes labeled approximate county evidence; it never supplies historical sale conditions.

A comparable whole-sale premium is (recorded comparable bid ÷ positive advertised baseline) − 1. Independently reconstructed baskets require every offered bid line, known quantity and price, and unambiguous line identities. Source-recorded totals remain distinguishable from reconstructed totals. Missing values do not become zero; incomplete or conflicting baskets and non-bid lines sold at appraisal are excluded or explicitly flagged. Winning-bid summaries remain separate from all recorded bids.

Condition features come from reproducible bounded rules applied to the exact matched offering’s prospectus, with source document, physical PDF page, excerpt, offering date, scope and extraction version. Negation, exceptions and payment-unit language can withhold a classification. Shared instructions do not automatically describe a specific sale. Not mentioned, not extracted, extraction failed and explicit absence are distinct states; later prospectus revisions cannot establish earlier restrictions.

Candidate conditions and distance bands are declared before comparison. Descriptive counts remain visible when coverage or group sizes are too small for a conclusion. Participation given an observed offering differs from premium among actual participants and wins among known outcomes. An archived offering does not establish that a buyer saw it or could bid. Effect sizes need both group denominators, uncertainty, period and exclusions; exploratory associations without independent later-period validation are labeled accordingly. Repeated lines and reoffers do not create independent tracts. No result infers motives, capacity, misconduct or future bids.

Mills & log yards starts with the Michigan DNR’s February 2025 wood products directory. It includes sawmills, pulp and paper mills, panel and veneer mills, and log yards; these are not all sawmills. Each business has its own profile, and map pins identify the type. Business locations are separate from buyer name groups. Only reviewed buyer connections link a mill to historical bidding; suggested name matches never silently merge companies.

Public business pages supply dated details and contacts. Each profile retains the source, excerpt, date checked and any unresolved or conflicting information in Sources & notes. A current web listing is not a tested phone number, confirmed email delivery or guarantee of operation. Historical directory contacts are explicitly identified as not currently confirmed. A timber procurement contact is shown only when the source states that role; a general office contact is labeled separately. A dated record of closure, acquisition or renaming is independent of DNR bidding activity.

Open map provides a stationary full-window view with wheel zoom, independently scrollable filters and lists, Escape to close and focus returned to its opening button. The same map retains its location between preview and full-window views. Map marks and accessible lists link to business profiles and exact opening evidence.

Precise mill proximity requires a verified mill association, suitable point accuracy, ownership and operating dates covering the opening, and a sale boundary with retained snapshot/date support. Where several mills qualify, the nearest eligible facility is identified. Straight-line geodesic distance is not haul distance, access, travel time or delivered cost. Locality-only points and logger offices cannot support a historical mill-procurement interpretation.

Present-day facility locations and the replace-all DNR boundary snapshot remain map context when historical validity is unknown. Missing dated coverage withholds the historical distance conclusion; unmapped buyer participation remains in the wider table. Managed land, roads and mapped water likewise cannot by themselves establish offered supply, legal truck access, a required crossing or an operational restriction.

On every sale that drew two or more bidders, the spread is the gap between the winning bidder’s total offer and the runner-up’s. A bidder’s total is their $/unit on each line times the DNR’s estimated volume for that line, summed over the sale — the same arithmetic the DNR uses to name a winner. Lines sold at the advertised value count equally for every bidder, and a bidder can win a sale while bidding lower on some lines: the total decides. Gap % is the winner’s total over the runner-up’s, minus one. Sales with a single bidder have no spread and are not in these figures. The spread describes what happened at the opening; it never says what anyone should have bid.

A state sale that draws no bids is usually offered again under the same sale number, usually at a lower advertised value. A chain is every bid opening held under one DNR sale number, in date order. Markdown is the advertised (minimum) total at the final offering against the first. A chain ends “sold” at the opening that drew a winning bid, and “unsold” if its last offering on record drew none. A chain marked “sold, offered again” sold and later returned to bid opening. A later offering follows an earlier winning bid; the record does not establish why. Openings whose sale number appears once are not chains.

A sale can carry two DNR offices. The unsold sheet files a sale under a field office; the office that held the bid opening can be a different one. Stumpage shows the opening office wherever the sale number joins the bid record, and names the sheet’s field office separately rather than picking one and hiding the other.

The bid record names only the DNR office; the prospectus names the county and the acres. Stumpage joins the two on sale number and bid date. $/acre is the winning total of the county’s sold sales over those sales’ prospectus acres; a sale that drew no bids counts as offered but is not in the $/acre figure. Sales before fiscal 2012 have no prospectus on file and are not in the county figures; a bid-opening result whose date matches no prospectus is left out rather than guessed.

The DNR’s Timber Sale Preparation bid tabulations (“Bid Results” workbooks, 2017 – present) list who bid to prepare each sale for the state. Stumpage archives every workbook. Workbook years are fiscal file labels; project bid dates retain their calendar dates. The 2021 workbook repeats the 2020 tabulations; repeats are counted once. A lump sum is the contractor’s total bid for the project; $/acre is the low bid over the project’s acres — a preparation cost, never a stumpage price. The low bid is what the tabulation shows; the DNR awards to the most advantageous bid, usually but not always the lowest. GNA marks Good Neighbor Authority work on federal land. Firm names are shown exactly as the tabulations spell them — the same firm appears under several spellings and is not merged.

Alongside the public DNR record, Stumpage accepts contributed results from private timber sales. This data is kept apart at every level — its own tables, its own pages — and it is always drawn in a different visual language: dashed lines, hollow markers, a “contributed” label. You will never see public and private numbers blended into one figure.

The private record is being built from accepted contributor reports. Regional results include only sales with the contributor’s separate publication agreement. A result may be based on one sale; its sale count is shown and is essential context, not evidence of a settled market price. Earlier submissions under the three-sale agreement remain excluded unless explicitly opted in.

Stumpage groups private results into the Upper Peninsula, Northern Lower Peninsula and Southern Michigan. The Northern Lower Peninsula includes Mason, Lake, Osceola, Clare, Gladwin and Arenac counties and the Lower Peninsula counties north of them; the remaining Lower Peninsula counties form the southern region. These are Stumpage reporting regions, not the DNR’s management regions. County outlines help identify a region; they do not locate a contributed sale. Names and precise locations are not published, but anonymity cannot be guaranteed.

For unit-price results, the published private figure is a median sale price, one sale one vote. Species, product, unit, log rule, grade and reported pricing basis stay visible. A one-sale median is that sale’s reported price. Known species prices can come from lump-sum or scaled sales; the sale type describes the selling arrangement, not whether species prices are available. New entries require volumes for every species row. If any species price is unknown, the contributor must enter the whole-sale total. Missing species prices are never inferred by allocating that total among species.

Reported whole-sale totals are summarized separately. Multiplying known species prices by their recorded volumes gives a calculated subtotal, which may differ from an actual payment. Calculated subtotals are not included in reported-total statistics. A sale can contribute both known species prices and a reported total, while counting only once in regional coverage.

Private oak prices are reported under the White Oak group or Red Oak group. A known subtype, such as Bur Oak or Black Oak, stays in the contributor’s entry and contributes to its parent group. Multiple subtype rows from one sale still count as one sale within the matching product, grade, unit and log-rule group. An unknown oak group is kept separate; Stumpage does not guess its classification.

A timber cruise estimates standing volume; measured harvested volume answers a different question. Species volumes may be estimated or scaled; the selling arrangement alone does not establish how they were measured. Existing entries may not record that basis; absence does not establish how the volume was measured. An amount per estimated unit should not be read as an amount per measured harvested unit. For an unchanged total payment, an overstated volume estimate lowers the implied price per unit. That arithmetic does not establish an error rate in DNR reports or private submissions. Units and log rules are not converted, and no blanket percentage adjustment is applied to DNR or private prices.